The present invention relates to a failure analysis technique for analyzing a failure of an electronic device or the like.
In fabrication of electronic devices such as a semiconductor memory typified by a dynamic random access memory (DRAM), a microprocessor, a semiconductor device such as a semiconductor laser, and a magnetic head, a high production yield is demanded.
Reduction in the product yield due to occurrence of a failure causes deterioration in profitability. Consequently, it is a big task to find a defect, a foreign matter, and poor processing as causes of a failure early and to take a countermeasure early. For example, at a manufacturing site of a semiconductor device, energies are put into early finding of a failure by a careful inspection and analysis of the cause of the failure. In a process of manufacturing actual electron devices using a wafer, a wafer in a process is inspected, the cause of an abnormal part such as a defect in a circuit pattern or a foreign matter is pursued, and a countermeasure is examined.
Usually, a high-resolution scanning electron microscope (hereinbelow, abbreviated as SEM) is used for observing a fine structure of a sample. As the packing density of a semiconductor increases, it is becoming impossible to observe an object with the resolution of the SEM, and a transmission electron microscope (hereinbelow, abbreviated as TEM) having a higher observation resolution is used in place of the SEM.
Preparation of a conventional sample for TEM accompanies a work of extracting a small piece from a sample substrate by cleavage, cutting, or the like. In the case where a sample substrate is a wafer, in most cases, the wafer has to be cut.
Recently, there is an example of using a processing method of irradiating a sample substrate with an ion beam and applying the action that particles constructing the sample substrate are discharged from the sample substrate by a sputtering action, that is, a focused ion beam (hereinbelow, abbreviated as FIB) process.
According to the method, first, a rectangular-shaped pellet having a thickness of sub-millimeters including an area to be observed is cut out from a sample substrate such as a wafer by using a dicer or the like. Subsequently, a part of the rectangular-shaped pellet is processed with an FIB into a thin film form, thereby obtaining a TEM sample. The feature of the FIB-processed sample for TEM observation is that a part of a sample piece is processed to a thin film having a thickness of about 100 nm so as to be observed by the TEM. Although the method enables a desired observation part to be positioned with precision of a micrometer level and observed, the wafer still has to be cut.
As described above, although the advantage of monitoring a result of a process during fabrication of a semiconductor device or the like is big from the viewpoint of yield management, a wafer is cut for preparing a sample and pieces of the wafer do not go to the following process but are discarded. Particularly, in recent years, the diameter of a wafer is increasing in order to lower the unit price of fabricating a semiconductor device. To be specific, the number of semiconductor devices which can be fabricated from one wafer is increased, thereby reducing the unit price. However, in other words, the price of a wafer increases and the number of semiconductor devices which are lost by discarding a wafer also increases. Therefore, the conventional inspection method including cutting of a wafer is very uneconomical.
Addressing the problem, there is a method capable of obtaining a sample without cutting a wafer. The method is disclosed in Japanese Patent Application Laid-Open No. 05-52721 (prior art 1).
According to the method, as shown in FIGS. 2(a) to 2(g), the posture of a specimen substrate 202 is kept so that the surface of the specimen substrate 202 is irradiated with an FIB 201 at the right angle, and a rectangular area in the surface of the specimen substrate 202 is scanned with the FIB 201, thereby forming a rectangular hole 207 having a required depth in the surface of the sample (FIG. 2(a)). After that, the specimen substrate 202 is tilted and a bottom hole 208 is formed. The tilt angle of the specimen substrate 202 is changed by a specimen stage (not shown) (FIG. 2(b)). The posture of the specimen substrate 202 is changed to set the specimen substrate 202 so that the surface of the specimen substrate 202 becomes perpendicular to the FIB 201 again, and a trench 209 is formed (FIG. 2 (c)). A manipulator (not shown) is driven to make the tip of a probe 203 at the end of the manipulator come into contact with a part to be extracted from the specimen substrate 202 (FIG. 2(d)). Subsequently, while supplying a deposition gas 205 from a gas nozzle 210, an area including the tip portion of the probe 203 is locally irradiated with the FIB 201, thereby forming an ion beam gas assisted deposition film 204. The separation part in the specimen substrate 202 and the tip of the probe 203 which are in contact with each other are connected by the ion beam assisted deposition layer 204 (FIG. 2(e)). The remaining part is cut with the FIB 201 (FIG. 2(f)) and a micro-sample 206 as an extracted sample is cut out from the specimen substrate 202. The cut-out micro-sample 206 is supported by the probe 203 connected (FIG. 2(g)).
The micro sample 206 is processed with the FIB 201 and an area to be observed is thinned, thereby obtaining a TEM sample (not shown). By introducing the micro-sample separated by the method into any of various analyzers, analysis can be conducted.
The above method is an example of extracting a micro-sample by a sample preparing apparatus and there is also a method of processing the shape of a sectional sample thin film, taking a specimen substrate from the sample preparing apparatus, and a sectional sample thin film is extracted by another mechanism in atmosphere. For example, a method is disclosed in “Material Research Society, Symposium Proceedings”, vol. 480, 1997, pp. 19 to 27 (prior art 2). Similarly, a method is disclosed in “Proceedings of the 22nd International Symposium for Testing and Failure Analysis, 18–22 November 1996”, pp. 199 to 205 (prior art 3).
As shown in FIG. 3(a), a section sample membrane 307 is formed while processing both sides of a target position on a wafer 308 in a stair shape with an FIB 301. Subsequently, by tilting a sample stage, the angle formed between the FIB 301 and the surface of the specimen is changed and the specimen substrate is irradiated with the FIB 301. As shown in FIG. 3(b), the peripheral portion of the sample membrane 307 is cut away by using the FIB 301 and the sample membrane 307 is separated from the wafer 308. The wafer 308 is taken out from an FIB apparatus, a glass stick is allowed to approach the process portion in the atmosphere, by using static electricity, the sample membrane 307 is absorbed by the glass stick and taken out from the wafer. The glass stick is moved to a mesh 309 and is either adsorbed by the mesh 309 by static electricity or disposed so that a surface to be processed faces a transparent adhesion member. In such a manner, without taking out the processed sample membrane in the system, even when most of the outer shape of the sample membrane is processed with an ion beam, by introducing the separated sample membrane into a TEM, analysis can be made.
A device manufacturing method in which a method similar to the prior art 1 for process management is disclosed in, for example, Japanese Patent Application Laid-Open No. 2000-156393 (prior art 4).
According to the method, process management is performed by a flow as shown in FIG. 4. A lot 401 is subjected to a process ml. After completion of the process ml, a predetermined number of wafers are selected as wafers 402 for inspection from the lot 401 and the other wafers 403 enter a standby mode. An area 404 for inspection in the selected wafer 402 for inspection is extracted as a micro-sample 405. The wafer 402 for inspection from which the micro-sample 405 is extracted is put together with the other wafers 403 again and the wafers as a lot 401A are subjected to the following process m2. The micro-sample 405 is processed so as to be used in one of analysis apparatus 406 and is transmitted to the analysis apparatus 406 where a target area in the micro-sample 405 is analyzed. A result of analysis is sent to a computer 407 and stored as a data base. The stored data base is transmitted as necessary via a communication path “h” to the process m1 or m2 and an instruction of a change in the process conditions or the like is given.
It is a big feature that a wafer is passed through paths a, b, c, and d from the process m1 to the process m2 and, during the paths, a micro sample for analysis is extracted. The number of sample substrates does not decrease for the inspection. The number of wafers in the lot 401 subjected to the process m1 and the number of wafers in the lot 401 subjected to the process m2 are the same. Consequently, there are no semiconductor devices which are lost due to cutting of the wafer. The total manufacture yield of semiconductor devices is increased and the manufacturing cost can be reduced.
In a failure analysis, when a failure mode is found by another tester such as a probe tester or an EB tester, a process causing the failure is clarified. The main target of the failure analysis in the invention is not only a failure existing only in a specific position in a wafer but also a failure existing in an entire face of a wafer or in an area of a certain range due to a process as a basic cause.
When a desired area is determined after a failure is detected by a test and a sample of observation and analysis is prepared by using means as employed in the prior arts 1, 2, and 3 in a procedure for failure analysis, the following problems remain.
Even if an abnormal part can be found in the observation sample prepared after device formation, there is a case that the process as a cause cannot be found in some cases. An example of the case will be described by referring to FIGS. 5(a) to 5(d).
FIG. 5(a) shows a cross section of a device on which a wiring process has been completed. In this example, formation of a metal line by a dual damascene process is shown. A metal line 501 is formed in a hole area for a line opened in an insulator layer 502. At this time point, the metal line 501 is normally formed. In the following process of forming a cap layer 503, heat treatment of about 300 to 400° C. is performed. In some cases, as shown in FIG. 5(b), a defect area 504 is formed in a connection hole part in the metal line 501 due to the heat treatment. Even in the case where no failure occurs in FIG. 5(b), there is a case that the defect area 504 occurs due to heat treatment of 400 to 500° C. at the time of forming an insulator layer 505 shown in FIG. 5(c).
However, in the case where after completion of a final process, for example, breaking of wire or a high-resistance part is found by a probe test, an area to be observed is determined, a section is formed by a method as described in the prior arts 1 to 3, and a wiring process is examined, although the defect area 504 as shown in FIG. 5(d) is observed, it is very difficult to clarify a process as a direct cause of the defect. Consequently, it is very important to clarify the cause from information which does not exert an influence by a later process.
An area to be monitored is preliminarily determined in the prior art 4, so that it is very effective for a process monitor for processing the area into a thin film or cross section to be observed. However, in order to use the area for failure analysis, the following problems occur.
In the failure analysis, an area for observation and analysis cannot be preliminarily specified. Consequently, if a micro-sample is extracted after completion of each process or a plurality of processes, and preliminarily processed as a TEM sample or the like, in the case where an area to be observed is determined in a later inspection and the area is different from the position of the prepared TEM sample, the possibility that the desired area has disappeared already is high, and the desired area cannot be observed. In the case of failure analysis, not only the position but also the direction of a face which is desired to be observed are also important. For example, in the case of a DRAM, there can be directions of sections parallel to and perpendicular-to a word line, a face parallel to the surface of a specimen, and the like. In consideration of combinations of the positions and directions, the possibility that the position of the prepared TEM specimen (or position of other cross sections for observation and analysis) coincides with the desired area of failure analysis is very low. Consequently, in the failure analysis, it is necessary to process the area for observation and analysis which is determined on the basis of failure data obtained after an inspection.